[erlang-questions] Constraint satisfaction problems

[Imre Palik]

During playing with project Euler I solved a few constraint satisfaction problems with erlang.  But whenever I code a solution, I have a really awkward feeling, that it should be possible to do better, but I don't know how.

I tend to code a backtracking search with constraint propagation before every step.  I model the constraint graph with digraph, but then the non-functional nature of the digraph package makes backtracking really awkward.

Is there any better way to solve constraint satisfaction problems?
What is the rationale behind those pesky side effects in digraph?

Thx

Imre